Going Upmarket: PLS Adoption and
MSME Retreat in Indonesian Islamic Banking

A quarterly bank-panel test of the authenticity–inclusion trade-off, 2021–2025

Ahmad Fatikhul Khasan

HAMPARAN Institute · Jember, Indonesia

ORCID 0000-0003-0209-834X · [email protected]

IIMEFC 2026 · Bank Indonesia Institute

The Puzzle

Two industry trends moving in opposite directions

Rising PLS

The authenticity signal

Industry-mean PLS share: 48% → 62%
Gain: +13.7 pp over 5 years
Both mudharabah and musyarakah rising

Falling MSME

The inclusion signal

Industry-mean MSME share: 24% → 22%
Loss: −2.3 pp over 5 years
Naïve bivariate: β = −0.52, p = 0.048

Scissor-like co-movement. Do these two trends reflect a structural trade-off — is Islamic banking “going upmarket”?

Start with a puzzle. Over the five years from 2021 through 2025, Indonesian Islamic banking has shown two simultaneous industry trends. The share of profit-and-loss-sharing contracts — the mudharabah and musyarakah akad, widely regarded as the more authentic expression of the Islamic risk-sharing principle — has risen by nearly fourteen percentage points, from 48 to 62 per cent. That is the authenticity signal. At the same time, the share of financing channelled to micro, small, and medium enterprises — the MSME segment, the inclusion mandate — has declined by 2.3 percentage points, from 24 to 22 per cent. Taken at face value, this scissor-like co-movement invites a simple interpretation. A naïve bivariate regression of MSME share on PLS share gives a coefficient of minus 0.52, significant at the five per cent level. The industry appears to be going upmarket — trading inclusion for authenticity. This paper asks whether that surface pattern survives rigorous panel analysis.

Contribution

Three contributions

1

First within-bank panel decomposition of PLS × MSME

242 bank-quarter observations covering all 13 Indonesian Bank Umum Syariah, 2021-Q1 to 2025-Q4
2

Theoretical extension of Stiglitz–Weiss to Islamic akad selection

Independence of authenticity and inclusion is the predicted equilibrium, not a puzzle
3

Digital-pivot diagnosis

Post-P2SK signal traced to two digital newcomers in transition, not to a steady-state structural relationship

Data infrastructure: 609,883 line items parsed from OJK Condensed Financial Statements · open replication.

Three contributions. First, methodologically. We assemble the first comprehensive within-bank panel decomposition of Islamic bank financing across both the akad-composition dimension and the MSME-segmentation dimension. 242 bank-quarter observations covering all thirteen Indonesian Bank Umum Syariah, spanning the full 2021-to-2025 window that includes the post-BSI-merger consolidation and the enactment of Undang-Undang Pengembangan dan Penguatan Sektor Keuangan in January 2023. Second, theoretically. We extend the classical Stiglitz–Weiss information-asymmetry framework to the Islamic akad-selection problem. We show that independence of authenticity and inclusion is the predicted equilibrium under monitoring-cost sorting, not a puzzle that needs explaining away. Third, empirically. We document a digital-pivot mechanism through which out-of-equilibrium transitions in young banks can generate mechanical within-bank correlations between strategic dimensions — a phenomenon we argue requires careful separation from steady-state structural relationships.

Where the paper sits

Literature: the within-Islamic akad decomposition is missing

Three strands frame the question:

▸ The murabahah syndrome critique

Khan (2010); Aggarwal & Yousef (2000) — debt-like akad dominance read as failure of Islamic ethics

▸ Islamic-vs-conventional comparative performance

Trinugroho et al.\ (2018); Rizvi et al.\ (2020); Yudaruddin (2023) — between-system, not within

▸ MSME financing in Islamic banks

Shaban et al.\ (2014); El-Komi & Croson (2013); Hachicha & Amar (2015) — MSME as single outcome

The gap → No comprehensive panel decomposition of Islamic bank financing across both akad composition and MSME segmentation on the full BUS universe.

The paper sits at the intersection of three literatures. First, the murabahah syndrome critique — anchored by Khan (2010) and by Aggarwal and Yousef (2000) — which reads the global dominance of debt-like akad as a failure of Islamic banks to honour their risk-sharing ethos. Second, the comparative performance literature — Trinugroho, Rizvi, Yudaruddin — which examines Islamic versus conventional banks but treats the within-Islamic akad composition as a fixed bank attribute. Third, the MSME financing literature — Shaban et al., El-Komi and Croson, Hachicha and Amar — which examines MSME exposure as a single outcome without decomposing by akad type. What is missing across all three is the joint panel decomposition of the two dimensions on the full universe of Indonesian Bank Umum Syariah. That is the gap this paper fills.

Data

A quarterly BUS panel from OJK CFS, 2021–2025

Bank Umum Syariah

242

bank-quarter observations

609,883

parsed line items

20 Q

2021-Q1 → 2025-Q4

Two focal variables

PLS share = (mudharabah + musyarakah) / total financing — from Posisi Keuangan
MSME share = Nasabah UMKM / (Nasabah UMKM + Bukan Nasabah UMKM) — from Kualitas Aset

Bank-level controls

Log assets, NPF gross, FDR, ROA (from Rasio Keuangan)
Bank age · ownership class (Persero, BPD, Swasta, Joint-Venture)
Sample straddles: BSI merger (Feb 2021), UU P2SK enactment (Jan 2023)

The data is built from the OJK Condensed Financial Statements micro-database, publicly accessible at cfs.ojk.go.id. The scraping and parsing pipeline yields 609,883 individual line items from the Posisi Keuangan, Kualitas Aset, and Rasio Keuangan reports. Aggregation produces a panel of 242 bank-quarter observations covering all thirteen active Bank Umum Syariah over twenty quarters, from 2021-Q1 through 2025-Q4. Two focal variables. PLS share is the ratio of mudharabah and musyarakah financing to total financing, computed from the Posisi Keuangan report. MSME share is the ratio of financing to Nasabah UMKM to total borrower financing, computed from the Kualitas Aset schedule. Bank-level controls come from the Rasio Keuangan report: log assets, non-performing financing, financing-to-deposit ratio, and return on assets. The sample straddles two important structural events — the BSI merger of February 2021 and the enactment of Undang-Undang Pengembangan dan Penguatan Sektor Keuangan in January 2023.

Two Focal Variables

Ratios computed directly from bank disclosures

PLS share — the authenticity dimension

PLS_i,t = ( mudharabah_i,t + musyarakah_i,t ) / Total Financing_i,t

Denominator sums 6 debt-like akad (murabahah, ijarah, qardh, salam, istishna, multijasa) plus the two PLS categories.

MSME share — the inclusion dimension

MSME_i,t = Nasabah UMKM_i,t / ( Nasabah UMKM_i,t + Bukan Nasabah UMKM_i,t )

MSME defined per POJK No.~11/2020 — annual revenue up to IDR~50~billion.

The two focal variables are ratios computed directly from the OJK bank disclosures. No estimation, no imputation, no proxy layer. PLS share is defined as the sum of mudharabah and musyarakah financing divided by total financing. The denominator sums the six debt-like akad — murabahah, ijarah, qardh, salam, istishna, multijasa — plus the two PLS categories. MSME share is defined as the ratio of financing to Nasabah UMKM to total borrower financing. The MSME category follows the OJK definition under POJK No. 11 of 2020: enterprises with annual revenue up to fifty billion rupiah. Both variables sit between zero and one and exhibit substantial cross-bank dispersion — from near-zero PLS at BTPN Syariah to near-saturation at Panin Dubai Syariah, and correspondingly from near-total MSME at BTPN to below fifteen per cent at Panin.

Empirical Strategy

Panel regression with two-way fixed effects

MSME_i,t = α_i + τ_t + β · PLS_i,t + γ' X_i,t + ε_i,t

α_i — bank fixed effects: absorb time-invariant business-model heterogeneity (ownership, region, founder identity)
τ_t — quarter fixed effects: absorb aggregate shocks (COVID, BSI merger window, UU P2SK enactment)
X_i,t — log assets, NPF, ROA, FDR (bank-level controls)
SE clustered at the bank level (allows within-bank serial correlation)
4 progressively-tighter specifications: pooled OLS → bank FE → two-way FE → full controls
Interpretation of β: within-bank, within-quarter conditional association — not a causal parameter

The empirical strategy is a panel regression with two-way fixed effects. MSME share is regressed on PLS share, with bank fixed effects, quarter fixed effects, and bank-level controls. The bank fixed effects — alpha-i — absorb all time-invariant heterogeneity in business model, ownership, region, and founder identity. The quarter fixed effects — tau-t — absorb aggregate shocks including COVID, the BSI merger window, and the UU P2SK enactment. Bank-level controls include log assets, NPF, ROA, and FDR. Standard errors are clustered at the bank level, which allows for within-bank serial correlation. We report four specifications of progressively tighter controls — pooled OLS, bank fixed effects only, two-way fixed effects, and the full specification with controls. Reading the coefficient across these four specifications is the substantive exercise of the paper. One caveat. The coefficient beta is a within-bank, within-quarter conditional association. It is not a causal parameter. Both PLS and MSME shares are jointly determined by bank strategy. What the regression identifies is whether within-bank deviations in one move systematically with within-bank deviations in the other.

Descriptive

Industry-mean trends — the “scissor”

Industry PLS and MSME trends, 2021Q1-2025Q4

What the picture shows

PLS share rises gradually — roughly monotonic
MSME share drifts down, most drop in 2024–2025
±1~SD bands remain wide throughout — heterogeneity persists, not dissipates
The industry-mean trend hides substantial bank-level variation

Section IV.C tests whether this cross-sectional pattern survives within-bank identification.

Figure 1 plots the cross-sectional means of PLS share in green and MSME share in red, for each quarter from 2021-Q1 through 2025-Q4, with shaded plus-or-minus one standard deviation bands across banks. Three features stand out. First, the PLS share rises gradually and roughly monotonically over the five-year window. Second, the MSME share drifts down, with most of the decline concentrated in 2024 and 2025. Third — and this is the important visual signal — the plus-or-minus one standard deviation bands are wide throughout and do not narrow over time. The heterogeneity across banks persists rather than dissipates. The scissor-like pattern in the industry means is what motivates the trade-off narrative. But the wide bands are the first hint that the industry-mean trend hides substantial bank-level variation. The formal test in the next slides will show that the surface correlation dissolves once that heterogeneity is partialled out.

Cross-Bank Heterogeneity

Four archetypes populate the PLS–MSME plane

Bank-level trajectories in PLS-MSME space

BSI — moderate “going upmarket”: PLS 36 → 47%, MSME 25 → 17%

Panin Dubai — saturated corporate-PLS: PLS > 96%, MSME < 15%

BTPN Syariah — near-pure debt-MSME: PLS ~ 0%, MSME > 91%

Aladin Syariah — explicit pivot: pure debt-MSME (2022) → PLS-corporate (2025) in 12 quarters

Different banks move in opposite directions along both axes. The industry mean masks this.

Figure 3 illustrates the heterogeneity of bank-level trajectories using four representative archetypes, each occupying a distinct corner of the PLS-MSME plane. BSI, the post-merger flagship, exhibits a moderate going-upmarket pattern — PLS rising from 36 to 47 per cent while MSME declines from 25 to 17 per cent. Panin Dubai Syariah operates in a saturated corporate-PLS regime throughout the entire window, with PLS share above 96 per cent and MSME share below 15 per cent. BTPN Syariah occupies the diametrically opposed corner — PLS share near zero, MSME share above 91 per cent throughout. This is a near-pure debt-MSME equilibrium. And Aladin Syariah, the digital newcomer, illustrates an explicit strategic pivot. From a pure-debt, pure-MSME configuration in 2022, the bank transitioned to a corporate-PLS-dominant profile by 2025 — a ninety-percentage-point swing on both dimensions within twelve quarters. Some banks move toward PLS while expanding MSME. Others swap dimensions entirely. Some hold their corner positions. The industry-mean trend in the previous slide is an artefact of averaging over these heterogeneous, sometimes opposing, bank-level paths.

Central Finding

The surface trade-off dissolves under panel identification

Specification	β (PLS)	SE	p	R²
(1) Pooled OLS	−0.521**	0.262	0.048	0.329
(2) Bank FE	−0.133**	0.056	0.018	0.038
(3) Bank FE + Time FE	−0.111	0.092	0.227	0.037
(4) Bank FE + Time FE + controls	+0.037	0.078	0.635	0.392

β = +0.037, p = 0.635 — statistically indistinguishable from zero.

PLS adoption and MSME serving are independent strategic dimensions once bank & time heterogeneity are partialled out.

This is the central finding of the paper. Table 4 reports the PLS-MSME relationship under four specifications of progressively tighter controls. Read the coefficient across the four columns. Column 1 — pooled OLS without fixed effects — yields the naïve bivariate coefficient of minus 0.521, significant at the five per cent level. This is the trade-off narrative in a single number. Column 2 adds bank fixed effects. The coefficient shrinks by a factor of four to minus 0.133, indicating that most of the surface correlation reflects time-invariant bank-level heterogeneity — the corner-bank positioning of Panin and BTPN we just saw in Figure 3. Column 3 adds quarter fixed effects to absorb aggregate shocks. The coefficient becomes statistically indistinguishable from zero — minus 0.111, p equals 0.227. Column 4 adds bank-level controls for size, NPF, ROA, and FDR. The coefficient flips sign to plus 0.037, with a p-value of 0.635. Within-bank deviations in PLS share are uncorrelated with within-bank deviations in MSME share once we control for aggregate shocks and bank fundamentals. The surface trade-off dissolves. The two dimensions are statistically independent. This is the central empirical result.

Robustness

Null holds across 7 sensitivity checks — save one

Spec	Variation	β	p
Base	Spec 4 of Table 4	+0.035	0.658
R1	Drop BSI	+0.032	0.695
R2	Drop Panin Dubai	+0.053	0.577
R3	Drop BTPN	+0.048	0.581
R4	Pre-P2SK (≤ 2022-Q4)	−0.214	0.578
R5	Post-P2SK (≥ 2023-Q1)	−0.211***	0.001
R6	+ lag PLS	+0.076	0.631
R7	+ lag MSME	−0.078	0.153

Reading the table

Composition checks (R1–R3): null holds
Dynamic checks (R6–R7): null holds
R5 post-P2SK anomaly: β = −0.211, p = 0.001

The post-P2SK signal → investigate in next slide.

The null finding is robust to seven sensitivity checks reported in Table 6. Composition checks first. Dropping BSI, Panin Dubai, or BTPN individually leaves the coefficient close to zero — plus 0.032, plus 0.053, plus 0.048 respectively — with p-values above 0.5. The null is not masking offsetting effects from any single corner bank. Dynamic checks. Adding a one-quarter lag of PLS share or of MSME share to absorb autoregressive dynamics preserves the null. Neither dynamic specification reaches conventional significance. One check does deviate. The post-P2SK sub-sample, spanning 2023-Q1 through 2025-Q4, yields a statistically significant negative coefficient — minus 0.211, p equals 0.001. This looks like the trade-off narrative re-emerging in the recent window. We investigate the source of this signal in the next slide, and the diagnosis will be important.

Interpretation

Independence is the predicted equilibrium under Stiglitz–Weiss

Two matching problems, not one

Akad selection: depends on borrower-specific monitoring cost γ. Corporate (low γ) → PLS. MSME (high γ) → debt-like.

Segmentation: depends on distribution network, regulatory mandate, target clientele.

Two decisions, two determinants — movements not mechanically linked.

Post-P2SK signal — digital-pivot diagnosis

Drop Aladin (in mid-transition): β = −0.150, p = 0.067

Drop Aladin + Nano: β = −0.147, p = 0.072

Mainstream 9 banks only: β = +0.158, p = 0.231

Post-P2SK signal is out-of-equilibrium transition, not steady state.

Why is the null the right answer? We interpret it through the classical Stiglitz–Weiss information-asymmetry framework, adapted to Islamic akad selection. The mechanism rests on borrower-specific monitoring cost gamma. PLS contracts require the bank to verify borrower revenue. For audited corporate borrowers with formal accounting, gamma is low. For opaque MSME borrowers without formal books, gamma is high. The optimal akad-borrower match segments naturally — corporate to PLS, MSME to debt-like. Critically, a bank's PLS share and MSME share are not two faces of a single allocation decision. They are two outputs of two distinct matching problems — the akad-selection problem (how to contract with each borrower) and the segmentation problem (which borrowers to serve). The first depends on monitoring technology. The second depends on distribution network and institutional mandate. Movements along the two dimensions are not mechanically linked. The post-P2SK signal is fully consistent with this frame. When we drop Aladin Syariah — the digital newcomer that executed an explicit business-model pivot from pure-debt-MSME to PLS-corporate in 2024 — the p-value moves above five per cent. Dropping both Aladin and Nano Syariah attenuates it further. Restricting to the nine mainstream incumbents reverses the sign entirely. The post-P2SK negative coefficient reflects mechanical correlation generated by out-of-equilibrium business-model transitions in two digital banks, not a steady-state structural relationship. Stiglitz-Weiss predicts steady-state independence; it does not need to predict transitional dynamics.

Policy Implications

Five levers to lower γ and close the PLS–MSME gap

1

Mandatory SAK EMKM adoption

Simplified accounting standards convert opaque MSME borrowers into verifiable counterparties
2

Digital monitoring infrastructure — QRIS, e-tax, e-commerce

Transactional data substitutes for traditional bookkeeping; POJK 21/2023 open-banking enables this
3

Promote hybrid akad — musyarakah mutanaqisah

Diminishing partnership reduces monitoring variance while preserving risk-sharing principle
4

Expand KUR Syariah with PLS-eligible variants

Third-party guarantee lowers γ without banks absorbing full monitoring cost themselves
5

MSME financial literacy programmes

Borrower-side γ reduction: train record-keeping + revenue documentation

If the policy goal is to achieve both higher PLS adoption and higher MSME serving jointly, the lever is neither contractual mandate nor segmentation quota. It is the underlying information cost gamma — the cost of verifying MSME revenue in a manner compatible with profit-sharing contracts. Five concrete levers. First, mandatory adoption of SAK EMKM — the simplified accounting standard for small enterprises. This converts presently opaque MSME applicants into verifiable counterparties. Second, digital monitoring infrastructure. QRIS transaction data, e-tax filings, and e-commerce histories provide independently verifiable signals of MSME revenue at near-zero marginal cost. The OJK Open Banking framework under POJK 21 of 2023 already collects this data; the bottleneck is regulatory permission for banks to use it in PLS underwriting. Third, promote hybrid contractual forms — particularly musyarakah mutanaqisah, the diminishing partnership. This reduces monitoring variance while preserving the risk-sharing principle of PLS. Fourth, expand KUR Syariah with PLS-eligible variants rather than the current debt-only structure. This transfers downside risk to the government balance sheet and lowers effective gamma. Fifth, MSME financial literacy programmes. This is the borrower-side margin — training MSME owners in record-keeping and revenue documentation improves their eligibility for revenue-contingent contracts. Each of these levers acts on the underlying information cost. None acts on the contract menu directly. That is the mechanism-design lesson from Stiglitz-Weiss.

Limitations · Next Steps

Honest limitations frame the follow-up agenda

Limitations

Associational, not causal — both variables are jointly determined by bank strategy
No jenis penggunaan breakdown in CFS public dataset (working-capital vs investment vs consumption)
UUS excluded — coarser MSME disclosure precludes consistent measurement

Next steps

Aceh Qanun LKS — sub-national natural experiment for causal identification
Firm-level bank-borrower matching for productivity/employment outcomes
KUR Syariah interaction with bank-level akad composition
Cross-country comparison (Malaysia, Pakistan, GCC)

Three honest limitations, each pointing to follow-up work. First, and most fundamentally, our identification strategy is associational, not causal. Both PLS and MSME shares are jointly determined by bank strategy. Our null finding is robust to this precisely because nulls cannot be amplified by omitted-variables bias of the conventional sort. But a future causal design — exploiting for example the staggered spin-off of UUS units into standalone BUS mandated by UU P2SK — would strengthen the inference. Second, the CFS public dataset does not disaggregate financing by jenis penggunaan — working-capital, investment, or consumption. This lacuna prevents observing whether PLS adoption is concentrated in working-capital lending, which is the segment with the most direct real-sector linkage. Third, we exclude the 24 Unit Usaha Syariah of conventional banks because their public reports use a coarser MSME breakdown. Future work as UUS disclosure matures should extend to the full Islamic banking universe. Four follow-up directions. The Aceh Qanun LKS conversion offers a sub-national natural experiment. Firm-level bank-borrower matching would identify productivity outcomes. KUR Syariah interaction merits direct empirical test. And cross-country comparison with Malaysia, Pakistan, and GCC systems would test external validity.

Conclusion

Authenticity and inclusion are independent
strategic dimensions in Indonesian Islamic banking —
not two faces of one trade-off.

Closing the dual gap requires investment in the information infrastructure
that lowers the verification cost of MSME revenue.

Terima kasih. I welcome your questions.

Ahmad Fatikhul Khasan · HAMPARAN Institute

[email protected] · ORCID 0000-0003-0209-834X

Going Upmarket: PLS Adoption andMSME Retreat in Indonesian Islamic Banking